Kimi K2.5/K2.6/K2.7-Code 1T — B200 vs B300
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and B300 (NVIDIA Blackwell) on Kimi K2.5/K2.6/K2.7-Code 1T. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Near the low end of the 35–169 tok/s/user interactivity band, at 68 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B200 runs 785 tok/s/GPU at $0.68/M tokens, B300 runs 933 at $0.70/M. B200 is 2% cheaper per token; B300 delivers 19% more tok/s/GPU.
Setting 102 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, B200 produces 415 tok/s/GPU ($1.30 per million tokens) and B300 produces 902 ($0.72). B300 is 81% cheaper per token; B300 delivers 117% more tok/s/GPU.
At 136 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, B200 delivers 207 tok/s/GPU at $2.73 per million tokens; B300 delivers 530 tok/s/GPU at $1.23. B300 is 123% cheaper per token; B300 delivers 156% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:784.9B300:933.1 | B200:414.9B300:902.2 | B200:206.7B300:530.3 |
| Cost ($/M tok) | B200:$0.684B300:$0.697 | B200:$1.303B300:$0.721 | B200:$2.733B300:$1.225 |
| tok/s/MW | B200:361718B300:430004 | B200:191190B300:415742 | B200:95276B300:244361 |
| Concurrency | B200:~24B300:~28 | B200:~8B300:~19 | B200:~3B300:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.